Feature points matching method based on improved A-KAZE algorithm

An improved A-KAZE algorithm is proposed in this paper. The nonlinear diffusion filtering strategy is used to improve the defect of the traditional Gauss kernel function in constructing the scale space, and the boundary of the target is preserved while the local adaptive filtering is realized. The characteristic points detected have scale invariance. Combining the principal direction of feature points and the improved BRIEF descriptor, the problem of rotation invariance in matching is solved, and the matching time is shortened in the form of binary coding. Compared with ORB, BRISK algorithm, the improved algorithm realizes the advantage combination of A-KAZE and BRIEF algorithm, and carries on the characteristic point matching experiment in indoor and outdoor environment, and compares it with that of BRIEF algorithm, and the result shows that the improved algorithm realizes the advantage combination of A-KAZE and brief algorithm. The high accuracy and high efficiency image feature point matching is realized.